A novel memory-saving technique of viterbi decoder for feedforward convolutional coding

Viterbi Algorithm(VA) has been invented for decades as the decoder of Convolution code. It helps reduce the computational complexity required to search for maximum likelihood signal sequence. Though VA could be used generally, people nowadays still try hard to improve it in two aspects. One is efficiency and the other is memory size. For the former, parallel hardware is introduced to block- process received signal sequence. The disadvantage of various methods in this category is mainly on the increase of hardware. As for the later problem, windowed TBM is developed to reduce the memory size which is used to store all the possible survival paths. The presence of windowed TBM is due to the finding of common tail. It is popularly applied and is considered as the minimal required memory length. However, for the general feedforward convolutional coders, there are some innate characteristics that we could use to improve both of the mentioned aspects. On one hand, efficiency could be slightly improved without adding extra hardware. On the other hand and most important of all, further reduction of the required memory length could be obtained to the extent that is several times smaller than length of the common tail. We name this technique as BW-TBM.